Resolution of singularities of arithmetical threefolds (Q2415312)

The problem of resolution of singularities -- solved by Hironaka about half a century ago over a field of characteristic 0 -- is still open in full generality. For arithmetical schemes \(\mathcal X\), the general case of \(\dim \mathcal{X} >2 \) remained unsolved until now. The next step is done here by the authors with their Theorem 1.1. Let \(\mathcal X\) be a reduced and separated Noetherian scheme which is quasi-excellent and of dimension at most 3. There exists a proper birational morphism \(\pi :\mathcal{X}'\to\mathcal{X}\) with the following properties: \begin{itemize} \item[(i)] \(\mathcal{X}'\) is everywhere regular; \item[(ii)] \(\pi\) induces an isomorphism \(\pi^{-1} (\mathrm{Reg} \ \mathcal{X}) \to\mathrm{Reg}\mathcal{X} \); \item[(iii)] \(\pi^{-1} (\mathrm{Sing } \ \mathcal{X})\) is a strict normal crossings divisor on \(\mathcal{X}'\). \end{itemize} If furthermore a finite affine covering \(\mathcal{X} = \mathcal{U}_1 \cup \mathcal{U}_2 \cup \dots \cup \mathcal{U}_n \) is specified, one may take \(\pi^{-1}(\mathcal{U}_i) \to \mathcal{U}_i\) projective, \(1\leq i \leq n\). A proper birational morphism \(\pi\) as above is said to be a resolution if it satisfies properties (i) and (ii) above. If additionally (iii) is satisfied, \(\pi\) is said to be a good resolution. It is pointed out that the construction of \(\pi\) is not given as a sequence of Hironaka-permissible blowing ups. The authors give evidence for situations, when this can be achieved proving a local version of the theorem which uses only Hironaka-permissible blowing ups. An answer to the question whether this could hold in general is referred to as \textit{widely open}. The following can be deduced from the theorem. Corollary 1.2. Let \(A\) be a reduced complete Noetherian local ring of dimension three. Then \(\mathcal{X} := \mathrm{Spec} \ A\) has a good resolution of singularities which is projective. Corollary 1.3. Let \(\mathcal O\) be an excellent Dedekind domain with quotient field \(F\) and \(\Sigma / F\) be a regular projective surface. There exists a proper and flat \(\mathcal O\)-scheme \(\mathcal X\) with generic fiber \(\mathcal{X}_F = \Sigma\) which is everywhere regular. In this paper the authors continue their earlier work on resolution of singularities of threefolds. In the first sections they start developping a more general approach to the problem for hypersurface singularities defined by a reduced polynomial \(h=X^p+f_1X^{p-1}+ \dots + f_p \in S[X], f_i\in S\) over an excellent regular local ring \(S\) of any dimension \(\geq 1\), where \(p:= \mathrm{char} (S/\mathbf{m}_S) >0\). It is supposed -- using the notations \(K:= Q(S)\), \(\mathcal{X}:= \mathrm{Spec} (S[X]/(h))\) and \(L\) for the total quotient ring of \(S[X]/(h)\) -- that \begin{itemize} \item[(i)] \(\mathrm{char} (K)=p\) and \(f_i =0\) for \(i<p\), or \item[(ii)] \(\mathcal{X}\) is \(G\)-invariant, where \(G:= \mathrm{Aut}_K(L) = \mathbb{Z}/(p)\). \end{itemize} Under these assumptions and for \(\mathrm{dim} (S)=3\), the main part of the article gives the following version of a resolution as stated in Theorem 1.5. Let \(\mu\) be a valuation of \(L\) which is centered in \(\mathbf{m}_S\). There exists a composition of local Hironaka-permissible blowing ups \((\mathcal{X}=: \mathcal{X}_0,x_0 ) \leftarrow (\mathcal{X}_1,x_1 ) \leftarrow \dots \leftarrow (\mathcal{X}_r,x_r ) \), where \(x_i\in \mathcal{X}_i\) is the center of \(\mu\), such that \((\mathcal{X}_r,x_r) \) is regular. Main combinatorial tool is a variant of Hironaka's characteristic polyhedron attached with the singularity. The inductive procedure is controlled by a numerical function which is different from the ``classical'' pair of multiplicity and slope function used for hypersurface singularities in residue characteristic 0. Applying an idea which can be traced back to Zariski, Theorem 1.5 implies the above Theorem 1.1.